In this section...
Overview
The Unable to show "metadata-from": No such page "_space_variables" performance metrics for simultaneous sessions for various call scenarios are summarized in the "Benchmarking Results" section of the following pages. Note that these numbers apply to our recommended CPU and memory configurations, and the results may vary depending on the actual CPUs used. Exceeding these simultaneous sessions metrics may result in poor call performance, including longer call setup times and/or poor voice quality. Sonus recommends maintaining call rates within the performance limits benchmarked below.
Server and VM Details
Server and VM Details
Configuration | Server Detail | VM Details |
Max Link | |||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Server Type | CPU Speed (GHz) | Processor sockets | RAM (GB) | Hyperthreading | vCPUs allocated | Memory (GB) | Disk Size (GB) | Network Interface Mode | ||
| Config-1 | HP DL380 Gen10 Gold 6140 | 2.3 | 2 | 256 | ACTIVE | 4 | 32 | 100 | Pass-Through Mode (Direct I/O) | 10 Gbps |
| Config-2 | 8 | Pass-Through Mode (Direct I/O) | ||||||||
| Config-3 | 20 | SR-IOV | ||||||||
| Config-4 | 32 | SR-IOV | ||||||||
Benchmarking Setup
The two test beds used to collect benchmarking data are depicted below.
The VM is hosted on a single NUMA node, where the NIC is connected.
Benchmarking Setup for SBC SWe Using SR-IOV interfaces
Benchmarking Setup for SBC SWe Configured in Pass-Through Mode (Direct I/O)
Performance Metrics
This section includes Sonus' benchmarking results using the following configurations with and without DoS. Please contact your Sonus representative if you are interested in benchmarking data for other call scenarios.
VMware Benchmarking Results G729a passthru Call Mix (x-code G729a->G711u + G729a passthru)
Call Scenario/Configuration Config Traffic Profile CPS/CHT Simultaneous
Sessions G729a passthru Config-1 (4 vCPU) standard_passthrough_profile 50/500 25,000 config-2 (8 vCPU) standard_passthrough_profile 50/283 41,150 config-2 (8 vCPU) standard_callmix_profile 150/147 22,050
KVM Hypervisor Benchmarking Results
Call Scenario/Configuration | Config | Traffic Profile | CPS/CHT | Simultaneous |
|---|---|---|---|---|
G729a passthru | Config-3 (20 vCPU) | standard_passthrough_profile | 600/77.5 | 46,500 |
x-code G729a->G711u | config-4 (32 vCPU) | standard_transcoding_profile | 180/31 | 5,580 |
G729a pass through with TLS/SRTP enabled on ingress leg | config-4 (32 vCPU) | standard_passthrough_profile | 500/88 | 44,000 |
G729a Direct Media | config-4 (32 vCPU) | standard_signaling_profile | 1000/60 | 60,000 |
SBC SWe Memory Sizes Based on Capacity Requirements
The following tables define the amount of VM memory required based on the call capacity expected and the configuration requirements.
Info
The
Unable to show "metadata-from": No such page "_space_variables"
uses the small configuration when the VM memory is less than 18 GB and the large configuration when the VM memory is 18 GB or greater. See SBC Provisioning Limits for an explanation of Unable to show "metadata-from": No such page "_space_variables"
small and large configurations.Note
The SBC SWe requires a minimum memory size of 10 GB.
Small Configuration
Number of Sessions | Memory Size Required |
|---|---|
5,000 | 10 GB |
18,000 | 14 GB |
| 24,000 | 16 GB |
28,000 | 17 GB |
Large Configuration
Number of Sessions | Memory Size Required |
|---|---|
12,000 | 18 GB |
22,000 | 20 GB |
Note
In the current implementation, the “send” and “receive” buffer size for TCP packets is restricted to 64 kilobytes. Due to this limitation, the SBC cannot handle traffic where the product of the TCP packet size (measured in kilobytes) and the processor rate (measured in cycles per second) exceeds 64. For example, if the packet size is 3 kilobytes and the processor rate is 25 cycles per second, the product is 75 - which is greater than the TCP buffer size. If the TCP buffer overflows, the SBC resets the connection. This is particularly important when the packet sizes are large, as in case of packets containing the INFO header that bloats the packet sizes.
Increasing RAM on Active SBC SWe Systems
Use the following procedure to increase RAM size on active (live) systems to meet the desired SBC SWe session capacity.
| Step | Action |
|---|---|
| 1 | Power down the designated Standby SBC. |
| 2 | Increase the RAM size of the VM system, and then power-on the Standby SBC. |
| 3 | Perform a manual switchover, and then power-down the previously Active SBC. Refer to the following pages for details: EMA: System - Admin CLI: Admin - CLI |
| 4 | Increase the RAM size of the VM system for the powered-down SBC, and then power it back up. |
| 5 | Manually switchover to revert the Active/Standby SBCs back to their original states prior to beginning this procedure. |
| 6 | Follow LSWU steps if you are upgrading to the next release. |
Overview
Content Tools